The Business Times
Garage logo
UOB logoBEST DIGITAL NEWS START-UP, WAN-IFRA 2019 GOLD AWARD

Alibaba rolls out open-source AI models to take on Meta's Llama 2

Published Thu, Aug 3, 2023 · 06:51 PM

Chinese tech giant Alibaba Group’s cloud computing unit released two open-source artificial intelligence (AI) models on Thursday in a bid to take on Meta Platforms.

The Hangzhou-based company said it had open-sourced two large language models (LLM), a type of AI model, named Qwen-7B and Qwen-7B-Chat on Thursday (Aug 3) in a press release.

Each model has seven billion parameters, which is often used to measure its strength. This marks the first time a big Chinese tech company has open-sourced its LLM.

It also comes after Meta unveiled a similar open-source model named Llama 2 last month, a move that has garnered widespread interest. Some analysts said that open-source models can chip away the current market dominance enjoyed by OpenAI, the developer of ChatGPT, and Google, whose AI models charge users expensive fees.

Alibaba Cloud announced its LLM named Tongyi Qianwen in April, which comes with multiple versions featuring different numbers of parameters. Alibaba said Qwen-7B and Qwen-7B-Chat are two small-sized versions of Tongyi Qianwen, aim at helping small and medium-sized businesses to start using AI.

The two models’ “code, model weights, and documentation will be freely accessible to academics, researchers and commercial institutions worldwide”, the company said.

A NEWSLETTER FOR YOU
Monday, 3.30 pm
Garage

The hottest news on all things startup and tech to kickstart your week.

However, companies with more than 100 million monthly active users will need to seek a licence from Alibaba before using the two models. Similarly, Meta’s Llama 2 also requires a licence from companies with more than 700 million users.

Alibaba said Tongyi Qianwen has versions with higher numbers of parameters which have not been open-sourced, but it did not say how many parameters its biggest model contains.

The biggest version of Meta’s Llama 2 features 70 billion parameters.

China has been trying to catch up to the US in the field of AI, as Beijing encourages Chinese companies to quickly develop home-grown and “controllable” AI models that can rival those developed by American companies.

Other than Alibaba, tech giants such as Tencent Holdings and Huawei have been aggressively developing their own AI models in recent months. Reuters

KEYWORDS IN THIS ARTICLE

READ MORE

BT is now on Telegram!

For daily updates on weekdays and specially selected content for the weekend. Subscribe to  t.me/BizTimes

Startups

SUPPORT SOUTH-EAST ASIA'S LEADING FINANCIAL DAILY

Get the latest coverage and full access to all BT premium content.

SUBSCRIBE NOW

Browse corporate subscription here